Create and Manage clusters via CAPA CAPI Provider

[richard-cox]

• Install / Setup provider
  • Handle install, setup and management of v2prov CAPI Providers + their UI Extensions #16539
  • AWS specific domains to proxy to need to be stored in the new ProxyEndpoint, this should be part of bootstraping the extension
    • Confirmed: These are shipped by default
• Step 1 - Select provider
  • UI must determine UI Extensions that provide appropriate v2prov capi providers
    • [RFE] Allow users with the clusters-create role to view capiprovider resource rancher#55167 will open up capiprovider to non-admins
  • TBD (UX) How to show the possible providers, do we show them alongside node driver options (even if they're the same provider)?
• Step 2 - Credentials
  • Currently rancher supports CC --> CAPA credential management via turtles and CAPI Provider
  • Spun out to Handle CloudCredentials and v2Prov CAPI CAPA clusters #17443 / https://github.com/SUSE/rancher-architecture/pull/58
• Step 3 - Create/Edit Cluster
  • Candidate to de-risk (start early with hacked other bits)
    • Updating the cluster create/edit component is a big thing
    • Mapping from machine config / showing new input fields for machine templates might be tricky
  • Fetching AWS specific node config
    • As before, /meta/proxy + reference to CC
  • Create infrastructure cluster object - AWSCluster
  • New node management area which CRUDs machine templates
    • Just like our process for node drivers inserting components we need same/similar for machine templates that provide a component
    • The output of these components is a machine template
    • We need to manage these before the user clicks save (same process as machine deployments). Create --> create resource, Edit --> replace resource, see below
    • The new machine templates are then references in the prov cluster
  • Editing machine templates is different than machine Deployments
    • Before - Just edit the machine deployment. Now - create a new machine template (clone?), save, assign new machine template to cluster
    • How are stale templates removed?
      • https://github.com/SUSE/rancher-architecture/pull/23#discussion_r2526533045
      • TBD (UX) UI will offer user choice to keep old template, by default this is unchecked
        • If checked UI will add a label to the old template marking it for GC
        • like machine configs there'll be a nightly cronjob that gets rid of tale ones
  • Created via the provisioning cluster
• Step 4 - View Cluster
  • Validate existing machine scale up / down actions work (including scale down on specific nodes)
• Step 5 - Delete Cluster
  • Delete the prov cluster
  • How are stale machine templates removed?
    • Similar to edit experience - https://github.com/SUSE/rancher-architecture/pull/23#discussion_r2558829434
    • TBD (UX) on delete of a cluster should we offer the user the chance to retain machine templates... or delete them when the cluster is deleted?
    • Should we remove Infrastructure cluster too?
• Documentation
  • Updates to allow UI Extensions support for above above should be documented in extension docs
• Other questions
  • Where will the extension live, built-in or separate repo?
  • Will extension be non-prime, given replacement to node drivers?